Facial recognition databases used by FBI have photos of about half of US adults without consent; algorithm for matches more likely to misidentify black people
Olivia Solon / The Guardian :
Context & Ripple Effects
This Guardian report extends a GAO audit that had already flagged privacy and accuracy concerns with the FBI's Next Generation Identification program, which held over 411 million photos including 173 million driver's license images at that point. The new findings add two sharper edges: the database covers roughly half of US adults who never consented, and the matching algorithm skews against Black faces.
First-order effects
- Millions of Americans whose only 'consent' was getting a driver's license are searchable by the FBI without ever being suspects, and Black individuals face elevated odds of being wrongly matched — meaning wrongful stops and investigations fall disproportionately on them.
Second-order effects
- State DMVs become the contested supply chain: the FBI's access runs through state license-photo databases, so the pressure lands on governors and motor vehicle agencies to decide whether to keep sharing, while civil-liberties groups gain concrete evidence linking scale and racial bias in one program.
Third-order effects
- The pattern holds and grows: by 2019 the FBI and ICE could reach 641 million photos across 21 states per internal documents and another GAO review, and a [[a:948939|study of 189 commercial algorithms confirmed systematic misidentification of people of color]]. Consent-free biometric collection from public records becomes the default posture of federal law enforcement, setting up the algorithmic-accountability fights that follow.
The trend: Law enforcement biometrics are scaling faster than consent norms or accuracy safeguards, turning routine government paperwork like driver's licenses into an implicit national face-search index.